JP2020123351A - Method and device for creating traffic scenario adapted to domain for virtual traveling environment for learning, testing, and verification of autonomous travel vehicle - Google Patents

Abstract

To provide a method and a device for creating an elaborate traffic scenario in a virtual travel environment on the basis of actual travel data.SOLUTION: A traffic scenario creation device inputs travel data into a situation analyzing device to extract travel environment information in a state where travel data has been acquired by using detailed traffic data corresponding to sequential traffic data collected from a sensor of a data collection vehicle of prior traffic data and an actual travel environment, which correspond to discrete traffic data extracted from existing travel video image by a video image base ADAS, and extracts vehicle state information for an own vehicle by inputting the travel data with a vehicle information extraction device to create a sequential traffic log with a travel sequence. Then, by inputting the sequential traffic log to a scenario enhancement network and using a critical event with a prior scenario enhancement network, the prior sequential traffic log is enhanced, the traffic scenario is verified, and the traffic scenario is mapped to a traffic simulation.SELECTED DRAWING: Figure 2

Description

The present invention relates to a virtual driving environment for learning an autonomous vehicle; more particularly, to a method and apparatus for generating a traffic scenario in the virtual driving environment for learning, testing and validating the autonomous vehicle.

Modern vehicles, especially automobiles, are increasingly offering autonomous driving and driving assistance systems such as blind spot monitors, automatic parking, and automatic navigation. However, testing and verifying autonomous driving and autonomous assistance systems is very complex and may require long-term road driving testing (eg, millions of hours and miles). Given that autonomous driving and updating of driving assistance systems require re-verification and individual verification is required depending on the vehicle type, testing and verification efforts can be doubled.

In addition, it is difficult to secure sufficient learning data for learning the autonomous traveling system and the traveling assistance system of the autonomous vehicle.

Therefore, there has been proposed a method for learning, testing, and verifying an autonomous traveling system and a traveling assistance system of an autonomous traveling vehicle that travels in a virtual traveling environment in which the actual traveling situation is virtualized.

In order to learn, test and verify the autonomous driving system and driving assistance system of an autonomous vehicle in a virtual driving environment, a scenario about surrounding information in the virtual driving environment should be provided.

However, in the conventional virtual traveling environment, since only scenarios for specific traffic situations are provided, there is a problem that it is not possible to provide various elaborate scenarios like the actual traveling environment.

The present invention aims to solve all the above-mentioned problems.

Another object of the present invention is to enable generation of various sophisticated scenarios in a virtual driving environment, such as an actual driving environment.

It is another object of the present invention to enable data augmentation to generate diverse and sophisticated traffic scenarios in a virtual driving environment.

The characteristic constitution of the present invention for achieving the above-mentioned object of the present invention and realizing the characteristic effects of the present invention described later is as follows.

According to an aspect of the present invention, in a method of generating at least one traffic scenario in at least one virtual traveling environment, (a) traveling data generated by referring to previous traffic data and detailed traffic data is acquired. In this state, the traffic scenario generation device inputs the traveling data to the situation analyzer, and the traveling condition information is extracted from the traveling data by the situation analyzer. With reference to the process of inputting to the extractor and having the vehicle information extractor extract the vehicle state information for the own vehicle (ego-vehicle) from the traveling data, and the traveling environment information and the vehicle state information, Performing a process of generating a sequential traffic log according to a driving sequence including a driving event; and (b) the traffic scenario generating device inputs the sequential traffic log to a scenario augmentation network, and By using at least one critical event as at least one condition with a scenario augmentation network, and augmenting the sequential traffic log so as to correspond to the event traveling environment of the critical event, the traffic scenario Generating the traffic scenario, verifying the traffic scenario with reference to preset reference traffic information, and mapping the traffic scenario to a traffic simulator in a virtual driving environment when the traffic scenario is determined to be valid. Performing the mapping process, the previous traffic data is obtained from at least one existing driving image including at least one of the driving events according to a vision-based ADAS (Advanced Driver Assistance System). The detailed traffic data is metadata corresponding to the extracted discrete traffic data, and the detailed traffic data are sequentially collected from one or more sensors of one or more data collection vehicles that are driven in an actual driving environment. A method characterized in that it is metadata corresponding to traffic data is provided.

As an example, a generator including an LSTM (long short-term memory) is used to obtain an initial traffic log of the sequential traffic logs and a next operation of the initial LSTM of the LSTMs. With the scenario set up as the state of the generator, the traffic scenario generator causes the generator to generate a predicted traffic log as the next operation corresponding to the traffic scenario, thereby causing the generator to generate the traffic scenario. And the predicted traffic log after the initial traffic log is enhanced to correspond to the event driving environment.

As one example, the traffic scenario generation device includes a process in which a discriminator learns the discriminator so as to distinguish between the sequential traffic log and the predicted traffic log, the sequential traffic log, and the predicted traffic. And a process of optimizing at least one parameter of the generator by a policy gradient using at least one reward of the discriminator generated by referring to the log. ..

As an example, if it is determined that the traffic scenario is not valid in the step (b), the traffic scenario generation device changes at least one of the conditions by changing the condition with reference to the verification process. The generated condition is generated, and the scenario increasing network is used to generate a new traffic scenario by increasing the sequential traffic log using the changed condition.

As one example, the transportation scenario generation device extracts the critical event from sequential data or extracts the critical event from a critical event database.

As one example, each of the sequential traffic logs, the number of at least one object, the position of the object, the number of at least one pedestrian, the position of the pedestrian, and at least one trajectory of the object, , Traffic signal change information.

As one example, the image-based ADAS uses the existing running image to detect an object, predict a distance, TSR (Traffic-sign Recognition), a roadway, a lane, and a TTC (time-to-). at least one of a collision prediction and a relative speed prediction.

As one example, the vehicle state information includes a speed of the host vehicle, an acceleration of the host vehicle, a steering wheel state of the host vehicle, a position of an accelerator pedal of the host vehicle, and a brake state of the host vehicle. And at least one of

According to another aspect of the present invention, in a traffic scenario generation device for generating at least one traffic scenario in at least one virtual driving environment, at least one memory storing instructions; and (I) previous traffic data, and details. In a state where the traveling data generated by referring to the traffic data is acquired, the traveling data is input to the situation analyzer, and the situation analyzer extracts the traveling environment information from the traveling data. A process, a process of inputting the traveling data to a vehicle information extractor, and having the vehicle information extractor extract vehicle state information for the own vehicle from the traveling data; and the traveling environment information and the vehicle state information. And (II) inputting the sequential traffic log to the scenario augmentation network, and performing the process of performing a process of generating a sequential traffic log according to a traveling sequence including traveling events, A process of generating the traffic scenario by using the at least one critical event as at least one condition to augment the sequential traffic log to correspond to an event driving environment of the critical event Performing a process of verifying the traffic scenario with reference to preset reference traffic information, and a process of mapping the traffic scenario to a traffic simulator in a virtual driving environment when the traffic scenario is determined to be valid. At least one processor configured to execute the instructions, the prior traffic data is extracted from at least one existing driving image including at least one driving event by the video-based ADAS. The detailed traffic data is metadata corresponding to sequential traffic data collected from one or more sensors of one or more data collection vehicles that are driven in an actual driving environment. There is provided a traffic scenario generation device characterized by being data.

As an example, a generator including an LSTM sets an initial traffic log of the sequential traffic logs as a state of the generator used to obtain a next operation of the initial LSTM of the LSTM. Then, the processor causes the generator to generate the predicted traffic log as a corresponding next operation with the scenario augmentation network to generate the traffic scenario, and the initial traffic log The subsequent predictive traffic log is augmented to correspond to the event driving environment.

As one example, the processor includes a process in which a discriminator trains the discriminator to distinguish between the sequential traffic log and the predicted traffic log, the sequential traffic log, the predicted traffic log, and Performing a process of optimizing at least one parameter of the generator according to a policy gradient using at least one reward of the discriminator generated by referring to the condition.

As an example, if the process (II) determines that the traffic scenario is not valid, the processor changes at least one of the modified conditions by referring to the verification process. And generate a new traffic scenario by sequentially increasing the traffic log using the changed condition with the scenario increasing network.

As one example, the processor extracts the critical event from sequential data or extracts the critical event from a critical event database.

As one example, each of the sequential traffic logs, the number of at least one object, the position of the object, the number of at least one pedestrian, the position of the pedestrian, and at least one trajectory of the object, , Traffic signal change information.

As one example, the image-based ADAS uses the existing running image to detect an object, predict a distance, TSR (Traffic-sign Recognition), a roadway, a lane, and a TTC (time-to-). at least one of a collision prediction and a relative speed prediction.

As one example, the vehicle state information includes a speed of the host vehicle, an acceleration of the host vehicle, a steering wheel state of the host vehicle, a position of an accelerator pedal of the host vehicle, and a brake state of the host vehicle. And at least one of

In addition, a computer-readable recording medium for recording a computer program for executing the method of the present invention is further provided.

According to the present invention, there is an effect that an elaborate traffic scenario in a virtual traveling environment can be generated based on actual traveling data.

Further, according to the present invention, there is an effect that various traffic scenarios can be generated in a virtual traveling environment by increasing discrete and sequential data for a specific traffic situation.

The following drawings, which are attached to explain the embodiments of the present invention, are merely some of the embodiments of the present invention, and have no ordinary knowledge in the technical field to which the present invention belongs. Those who have (hereinafter, "regular technician") can obtain other drawings based on this drawing without performing the inventive work.
FIG. 1 is a diagram schematically illustrating a traffic scenario generation device that generates at least one traffic scenario in a virtual driving environment according to an exemplary embodiment of the present invention. FIG. 2 is a diagram schematically illustrating a scenario generation method for generating the traffic scenario in the virtual driving environment according to an exemplary embodiment of the present invention. FIG. 3 is a diagram schematically illustrating a process of adding at least one condition in the method of generating the traffic scenario in the virtual driving environment according to an embodiment of the present invention. FIG. 4 is a diagram schematically illustrating a process of learning a scenario augmentation network in the method of generating the traffic scenario in the virtual driving environment according to an embodiment of the present invention. FIG. 5 is a diagram schematically illustrating a process of learning a scenario augmentation network in the method of generating the traffic scenario in the virtual driving environment according to an embodiment of the present invention.

DETAILED DESCRIPTION The following detailed description of the invention refers to the accompanying drawings that illustrate, by way of illustration, specific embodiments in which the invention may be practiced, in order to clarify each object, technical solution, and advantage of the invention. These examples are described in sufficient detail to enable a person of ordinary skill in the art to practice the invention.

Also, the word "comprising" and variations thereof throughout the detailed description of the invention and the claims are not intended to exclude other technical features, additions, components or steps. .. Other objects, advantages and features of the invention will be apparent to one of ordinary skill in the art, in part, from the description and partly from the practice of the invention. The following illustrations and drawings are provided by way of illustration and are not intended to limit the invention.

Furthermore, the invention covers all possible combinations of the examples presented herein. It should be understood that the various embodiments of the invention differ from each other, but need not be mutually exclusive. For example, the particular shapes, structures and characteristics described herein are, in connection with one example, without departing from the spirit and scope of the invention and may be implemented in other embodiments. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the detailed description set forth below is not to be taken in a limiting sense, and the scope of the present invention, when properly described, includes all ranges equivalent to what the claims claim It is limited only by the claims appended hereto. In the drawings, like reference numerals refer to the same or similar features across several aspects.

The various images referred to in the present invention may include images related to paved or unpaved roads, where objects that may appear in the road environment (eg, cars, people, animals, plants, objects, buildings, airplanes, A flying object such as a drone and other obstacles can be assumed, but the invention is not limited to this. The various images referred to in the present invention are images not related to roads (for example, unpaved roads). Images associated with roads, alleys, vacant lots, seas, lakes, rivers, mountains, forests, deserts, sky, indoors), in which case unpaved roads, alleys, vacant lots, seas, lakes, rivers, mountains, forests , Deserts, sky, objects that may appear in indoor environments (eg, cars, people, animals, plants, objects, buildings, air vehicles such as airplanes and drones, and other obstacles), but not necessarily It is not limited.

The titles and abstracts of the present invention provided herein are for convenience only and are not intended to limit the scope of the examples or to analyze the meaning of the examples.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those having ordinary skill in the art to which the present invention pertains can easily implement the present invention. ..

FIG. 1 is a diagram schematically illustrating a traffic scenario generation device that generates at least one traffic scenario in a virtual driving environment according to an exemplary embodiment of the present invention. Referring to FIG. 1, the traffic scenario generation device 100 includes a memory 110 storing instructions for generating the traffic scenario in the virtual driving environment, and a virtual memory corresponding to the instructions stored in the memory 110. A processor 120 may be included to perform the process of generating the traffic scenario in a driving environment.

Specifically, the traffic scenario generator 100 is typically a computing device (eg, a computer processor, memory, storage, input and output device, or other device that may include existing computing components; Electronic communication devices such as routers, switches, etc.; electronic information storage systems such as network attached storage (NAS) and storage area networks (SAN)), and computer software (ie, having the computing device to function in a specific manner) Can be used to achieve desired system performance.

The processor of the computing device may include an MPU (Micro Processing Unit) or a CPU (Central Processing Unit), a cache memory (Cache Memory), a data bus (Data Bus), or the like. In addition, the computing device may further include an operating system and a software configuration of an application performing a specific purpose.

However, it is not excluded that the computing device 100 includes an integrated processor that is an integrated form of a medium, processor, memory or other computing component for carrying out the present invention.

A method of generating the traffic scenario in the virtual traveling environment by using the traffic scenario generation device 100 according to the embodiment of the present invention configured as described above will be described with reference to FIG.

First, the traffic scenario generation device 100 may support the previous traffic data and the detailed traffic data, or may be supported by another device.

At this time, the previous traffic data may be data that is difficult to be acquired in an actual driving environment such as a vehicle accident, and at least one driving is performed by using the current computer image technology base ADAS (Advanced Driver Assistance System). The metadata may be extracted from at least one existing driving image including an event. On the other hand, the image-based ADAS uses the existing running image to detect an object, predict a distance, TSR (Traffic-sign Recognition), a roadway, a lane, and a TTC (time-to-collision). And at least one of relative velocity prediction.

The detailed traffic data may be metadata collected from one or more sensors of a vehicle traveling in an actual traveling environment, such as a lidar (LiDAR), a radar (Radar), a camera, and a GPS device.

Meanwhile, the previous traffic data may be discrete traffic data, and the detailed traffic data may be sequential traffic data.

Next, the traffic scenario generation device 100 may acquire traveling data by referring to the previous traffic data and the detailed traffic data. As an example, the travel data may be the travel data associated with the vehicle, such as information obtained from the GPS device, the rider, the radar, the camera, etc. corresponding to an input sensor of the vehicle.

Then, the traffic scenario generation device 100 inputs the travel data to the situation analyzer 130 and causes the situation analyzer 130 to extract travel environment information from the travel data, and the travel data. To the vehicle information extractor 140, and the vehicle information extractor 140 may perform a process of extracting vehicle state information for the own vehicle (ego-vehicle) from the travel data. At that time, as is generally known to a person skilled in the art, the own vehicle may mean a target vehicle, for example, a vehicle which is driven in the virtual traveling environment according to the present invention.

As an example, the situation analyzer 130 includes object detection information, distance prediction information, TSR (Traffic-sign Recognition) information, roadway detection information, lane detection information, TTC (time-to-collision) prediction information, relative speed prediction information. At least a part of the above can be acquired from the travel data.

In addition, the vehicle information extractor 140 includes the speed of the host vehicle, the acceleration of the host vehicle, the steering wheel state of the host vehicle, the position of the accelerator pedal of the host vehicle, and the brake state of the host vehicle. At least one of these can be extracted.

Next, the traffic scenario generation device 100 refers to the traveling environment information from the situation analyzer 130 and the vehicle state information from the vehicle information extractor 140 to refer to a traveling sequence including the traveling event ( Sequential traffic logs can be generated.

As an example, the traffic scenario generation device 100 generates an image frame by matching the traveling environment information acquired by the condition analyzer 130 with the vehicle state information acquired from the vehicle information extractor 140. be able to. Here, the image frame may be discrete and sequential data, which may be the traveling data before and after a specific event such as a tunnel or bridge entry event, a lane change event, a tunnel or bridge passage event, and the like.

Then, the traffic scenario generation device 100 post-processes the travel event included in the discrete and sequential data to thereby generate one or more sequential discrete processed data. Traffic logs can be generated.

At this time, each of the sequential traffic logs includes the number of at least one object, the position of the object, the number of at least one pedestrian, the position of the pedestrian, at least one trajectory of the object, and the traffic. Signal change information.

Then, the traffic scenario generation device 100 may sequentially convert the sequential traffic log into metadata.

That is, the traffic scenario generation device 100 encodes the data extracted by the condition analyzer 130 and the vehicle information extractor 140 included in the sequential traffic log by using at least one sequential meta-converter. , At least one traffic simulator can identify the encoded data.

Next, the traffic scenario generation device 100 inputs the sequential traffic log to the scenario augmentation network 150 and uses the scenario augmentation network 150 to utilize at least one critical event as at least one condition to generate the sequential traffic. The process of generating the traffic scenario may be performed by augmenting the sequential traffic log so that the log corresponds to the event driving environment of the critical event.

As an example, LSTM (long short-term)
state in which the generator 151 including the memory sets the initial traffic log of the sequential traffic logs as the state of the generator 151 used to obtain the next operation of the initial LSTM of the LSTM. Then, the scenario augmentation network 150 causes the generator 151 to generate the traffic scenario by generating one or more predicted traffic logs as the next operation corresponding to the predicted traffic log. The predicted traffic log after the log may be augmented to correspond to the event driving environment.

Referring to FIG. 3, the scenario augmentation network 150 inputs the at least one latent code z and the at least one critical event y to the generator 151 so that the generator 151 has The predicted traffic log corresponding to the initial traffic log may be generated so that the predicted traffic log corresponds to the critical event in the traveling environment.

At this time, the critical event may be sequentially extracted from the data, or a specific critical event may be retrieved from the critical event database 160 that stores various critical events.

As an example, the critical event may be a specific driving event among driving events such as a sudden stop in an emergency situation, a traffic light change at an intersection, and the like, in which the green light is changed to a red light at the intersection. Various conditions may be set, and the generator 151 may generate various traffic scenarios such as the actual driving conditions according to the critical event and the various conditions.

On the other hand, in the scenario enhancement network 150, the generator 151 that generates the traffic scenario and the discriminator 152 that evaluates the traffic scenario are learned in parallel and competitively, and the generator discriminates the discrimination. It can be configured to generate so sophisticated data that it can be tricked into trickery.

The scenario augmentation network 150 includes at least one state and at least one operation, is easy to model the discrete data, and is partially stochastic policy based on the reward. ) Can be acquired, the generator 151 can be trained by using a policy gradient, which is a policy-based reinforcement learning technique. Also, the Monte Carlo policy gradient among the policy gradient techniques may be used. In this case, the Monte Cal policy gradient is (i) a roll-out policy that is one of the networks that can predict the next state from the current state, and (ii) one of efficient non-exhaustive search algorithms. MCTS (Monte Carlo Tree Search) is a learning method for generating the probabilistic policy. As an example, the state may be set as a token, the action may be set as an action of giving a next token, and an agent may be set as the generator 151. The label determined by the discriminator 152 may be modeled as the reward to train the generator 151. In this case, the generator 151 may use at least one LSTM in the RNN (Recurring Neutral Network) without any vanishing gradient problem, and the discriminator 152 may use the CNN. ..

That is, referring to FIG. 4 and FIG. 5, the scenario augmentation network 150 inputs the predicted traffic log and the sequential traffic log generated by the generator 151 to the discriminator 152, thereby determining the discriminator. 152, it is possible to determine whether the input data of this is actual data or false data, and use one or more outputs of the discriminator 152 to train the discriminator 152. The discriminator 152 may distinguish the sequential traffic log and the predicted traffic log.

In addition, after learning the discriminator 152, the generator 151 including the LSTM uses the initial traffic log of the sequential traffic log to obtain the next operation of the initial LSTM of the LSTM. In the state set as the state of the generator 151, the scenario augmentation network 150 causes the generator 151 to generate the predicted traffic log as the next operation corresponding thereto, and the sequential traffic log and the prediction are generated. By inputting the traffic log and the condition to the discriminator 152, the discriminator 152 indicates whether or not the predicted traffic log is generated in accordance with the sequential traffic log and the condition. Rewards can be output. Then, the scenario enhancement network 150 may train the generator 151 by referring to the policy gradient using the reward.

Next, the traffic scenario generation device 100 may perform a process of verifying the traffic scenario with reference to preset reference traffic information.

At this time, the traffic scenario generation device 100 may determine whether or not the traffic scenario is valid by comparing the minimum traffic traffic information set in advance with the traffic scenario. As an example, assuming that the traffic scenario includes a highway approach event, all events in the traffic scenario occur during traveling on the highway, and referring to the traffic regulation database, the traffic on the highway is If there is the traffic rule information that the signal is not allowed, the event related to the traffic signal may not be valid.

When it is determined that the generated traffic scenario is valid, the traffic scenario generation device 100 can match the traffic scenario with the traffic simulator in the virtual traveling environment.

On the other hand, when it is determined that the traffic scenario is not valid, the traffic scenario generation device 100 changes at least one changed condition by changing the condition with reference to the verification process as described above. A new traffic scenario can be generated by performing the process of generating, and having the scenario augmentation network 150 augment the sequential traffic log using the changed conditions.

As described above, the present invention generates static data for a specific traffic situation and the sequential data, and generates the traffic scenario for a target vehicle, a pedestrian, etc., which is peripheral information in the virtual traveling environment. The generated traffic scenario can be mapped to the traffic simulator in the virtual traveling environment, and the generated traffic scenario for virtual traveling can be utilized.

Also, the embodiments of the present invention described above may be embodied in the form of program command words that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include a program command word, a data file, a data structure, etc., alone or in combination. The program command words recorded on the computer-readable recording medium may be those specially designed and constructed for the present invention, or those known and available to those skilled in the computer software field. Good. Examples of computer-readable recording media are magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magnetic-optical media such as floppy disks. Included are magneto-optical media and hardware devices specially configured to store and execute program command words such as ROM, RAM, flash memory, and the like. Examples of program instruction words include not only machine language codes such as those produced by a compiler, but also high level language codes that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules to perform the process of the present invention, and vice versa.

The present invention has been described above with reference to specific examples such as specific components and limited examples and drawings, which are provided to help a more general understanding of the present invention. However, the present invention is not limited to the above-mentioned embodiments, and various modifications and variations can be made by those skilled in the art to which the present invention pertains.
Therefore, the idea of the present invention should not be limited to the above-described embodiments, and is not limited to the scope of the claims to be described later, and all modifications equivalent or equivalent to the scope of the claims. Can be said to belong to the scope of the idea of the present invention.

Claims (16)

In a method of generating at least one traffic scenario in at least one virtual driving environment,
(A) With the travel data generated by referring to the previous traffic data and the detailed traffic data being acquired, the traffic scenario generation device inputs the travel data to the situation analyzer, and the situation analyzer And a process for extracting traveling environment information from the traveling data, inputting the traveling data to a vehicle information extractor, and causing the vehicle information extractor to extract vehicle state information for the own vehicle from the traveling data. Performing a process, and a process of generating a sequential traffic log according to a driving sequence including a driving event, with reference to the driving environment information and the vehicle state information; and (b) the traffic scenario generation device performs the sequential processing. The traffic log is input to the scenario augmentation network, the scenario augmentation network is used to utilize at least one critical event as at least one condition, and the sequential traffic log is adapted to correspond to the event traveling environment of the critical event. By sequentially increasing the traffic log, a process of generating the traffic scenario, a process of verifying the traffic scenario with reference to preset reference traffic information, and the traffic scenario is determined to be effective, Performing a process of mapping the traffic scenario to a traffic simulator of a virtual driving environment;
The previous traffic data is metadata corresponding to discrete traffic data extracted from at least one existing driving video including at least one driving event by the video-based ADAS, and the detailed traffic data is actually The method is characterized in that it is metadata corresponding to sequential traffic data collected from one or more sensors of one or more data collection vehicles that are driven in the driving environment. With the generator including the LSTM setting the initial traffic log of the sequential traffic logs as the state of the generator used to obtain the next operation of the initial LSTM of the LSTM, The traffic scenario generation device causes the generator to generate the traffic scenario by generating a predicted traffic log as the next operation corresponding to the predicted traffic log with the scenario augmentation network. The method of claim 1, wherein the predictive traffic log is augmented to correspond to the event driving environment. In the traffic scenario generation device, a discriminator learns the discriminator so as to distinguish between the sequential traffic log and the predicted traffic log, the sequential traffic log, the predicted traffic log, and the condition. The method of optimizing at least one parameter of the generator by a policy gradient utilizing at least one reward of the discriminator generated with reference to. .. In the step (b),
When it is determined that the traffic scenario is not valid, the traffic scenario generation device generates at least one changed condition by changing the condition with reference to the verification process, The method according to claim 1, wherein a new traffic scenario is generated by augmenting the sequential traffic log using the changed condition. The method according to claim 1, wherein the traffic scenario generation device extracts the critical event from sequential data or extracts the critical event from a critical event database. Each of the sequential traffic logs includes at least one object number, the position of the object, the number of at least one pedestrian, the position of the pedestrian, at least one trajectory of the object, and traffic signal change information. The method of claim 1, comprising: The image-based ADAS performs at least one of object detection, distance prediction, TSR, roadway detection, lane detection, TTC prediction, and relative speed prediction using the existing running video. The method according to claim 1, wherein: The vehicle state information is at least one of a speed of the own vehicle, an acceleration of the own vehicle, a steering wheel state of the own vehicle, a position of an accelerator pedal of the own vehicle, and a brake state of the own vehicle. The method of claim 1 including one. In a traffic scenario generation device for generating at least one traffic scenario in at least one virtual traveling environment,
At least one memory storing instructions; and (I) inputting the travel data into the situation analyzer with the travel data generated by referring to the previous traffic data and the detailed traffic data, A process for extracting traveling environment information from the traveling data by the situation analyzer, inputting the traveling data to a vehicle information extractor, and using the vehicle information extractor, a vehicle state for the own vehicle from the traveling data. A process for extracting information, a process for performing a process for generating a sequential traffic log by a driving sequence including a driving event with reference to the driving environment information and the vehicle state information, and (II) the sequential The traffic log is input to the scenario augmentation network, the scenario augmentation network is used to utilize at least one critical event as at least one condition, and the sequential traffic log is adapted to correspond to the event traveling environment of the critical event. By sequentially increasing the traffic log, a process of generating the traffic scenario, a process of verifying the traffic scenario with reference to preset reference traffic information, and the traffic scenario is determined to be effective, At least one processor configured to execute the instructions for performing a process of mapping the traffic scenario to a traffic simulator of a virtual driving environment;
The previous traffic data is metadata corresponding to discrete traffic data extracted from at least one existing driving video including at least one driving event by the video-based ADAS, and the detailed traffic data is actually A traffic scenario generation device characterized by being metadata corresponding to sequential traffic data collected from one or more sensors of one or more data collection vehicles that are driven in the driving environment. With the generator including the LSTM setting the initial traffic log of the sequential traffic logs as the state of the generator used to obtain the next operation of the initial LSTM of the LSTM, The processor causes the generator to generate the traffic scenario by causing the generator to generate a predicted traffic log as a corresponding next operation with the scenario augmentation network, and the predicted traffic after the initial traffic log. The traffic scenario generation device according to claim 9, wherein the log is augmented to correspond to the event driving environment. The processor refers to a process in which a discriminator learns the discriminator so as to distinguish between the sequential traffic log and the predicted traffic log, the sequential traffic log, the predicted traffic log, and the condition. The process of optimizing at least one parameter of the generator according to a policy gradient using at least one reward of the discriminator generated by the traffic scenario generation device according to claim 10. .. In the process (II),
If the traffic scenario is determined to be invalid, the processor generates at least one modified condition by modifying the condition with reference to the verifying process, and the processor augments the modified condition with the scenario augmentation network. The traffic scenario generation device according to claim 9, wherein a new traffic scenario is generated by increasing the sequential traffic log using the specified condition. The traffic scenario generation device according to claim 9, wherein the processor extracts the critical event from sequential data or extracts the critical event from a critical event database. Each of the sequential traffic logs includes at least one object number, the position of the object, the number of at least one pedestrian, the position of the pedestrian, at least one trajectory of the object, and traffic signal change information. The traffic scenario generation device according to claim 9, comprising: The image-based ADAS performs at least one of object detection, distance prediction, TSR, roadway detection, lane detection, TTC prediction, and relative speed prediction using the existing running video. The traffic scenario generation device according to claim 9, wherein The vehicle state information is at least one of a speed of the own vehicle, an acceleration of the own vehicle, a steering wheel state of the own vehicle, a position of an accelerator pedal of the own vehicle, and a brake state of the own vehicle. The traffic scenario generation device according to claim 9, wherein the traffic scenario generation device includes one.

Images